Never "edit" your videos the same way again
Gemini Omni is the video AI game-changer everyone’s been waiting for. I’ve been testing it the last 24 hours. It’s not just another text-to-video model. It’s a different category.
It takes video as input
01 / What’s new
That’s the headline. Most video AI takes a text prompt and renders a clip. Omni takes video, images, audio, and text as direct inputs. You literally talk to your footage and tell it what to change.
I dropped in a single talking-head clip of myself, ran a few different prompts, and watched the model rewrite the scene without re-rendering my face from scratch. The identity stays. The motion stays. The thing you wanted to change, changes.
The three things that make it special
02 / Why it’s different
After a day of pushing it, these are the unlocks that actually matter:
- // 01 multimodal in: Video + image + audio + text as direct inputs. Edit by conversation, not by timeline.
- // 02 memory: It remembers what you asked for across edits. Iterate without re-explaining the whole scene.
- // 03 motion + text: Hyper-smooth motion. Text inside the frame renders flawlessly and tracks with the surface it sits on.
Older video models always glitch out, warp, or render text as unreadable mush. Omni puts text on a moving surface and keeps it sharp the whole way through. That alone changes what you can ship.
Free for everyone, today
03 / The release
Yesterday this was gated behind a waitlist and a paid Google AI plan. Today Google flipped the switch globally.
Free tier · Now live
3 generations / day
No card, no waitlist, available worldwide. Want more? Upgrade inside Google Flow for monthly credits (AI Plus / Pro / Ultra).
You can use it inside Google Flow or directly in the native Gemini app. Same model, two surfaces.
Honestly? I’d try this today
Open Google Flow, drop in a clip, talk to it. The creative ceiling just moved.